Prediction Markets: A Better Alternative

by Thomas Peick and Hubertus Hofkirchner -- Vienna, 17 Feb 2017

This is an english translation of an original article in Research & Results magazine from February 17th, 2017.

The success of AfD (“Alternative for Germany”) party in three German regional elections, the British referendum to leave the EU, the election victory of Donald Trump in the USA: pollsters worldwide were not able to predict these results. Hubertus Hofkirchner, Chief Futurist and CEO of Prediki, which operates the global prediction platform prediki.com, talks about the advantages of modern political stock markets.

Political stock markets have been known to produce significantly better quantitative predictions compared to telephone and online surveys. Innovative prediction market technology can now also decode the reasons and motivations of voters. It offers campaign managers a much improved alternative compared to focus groups and in-depth interviews.

Massive forecast errors

2016 was a horror year for German pollsters. The nemesis of traditional agencies goes by the abbreviation AfD. None of the four leading German polling institutes were able to estimate the vote share of the new party even remotely accurately.

The smallest prediction error - when looking at the final survey immediately before the elections - was 2.1 percentage points, the highest error was even at 6.2 percentage points. On average the prediction error was 4.2 percentage points. The last Sunday questions came in with a wide variance of the official final results. Predictions for the elections in Saxony-Anhalt were particularly inaccurate. The average prediction error was 5.7 percentage points, which is far beyond the theoretical confidence intervals.

Once more the virtual political stock markets fared significantly better than their traditional competitors. The prediction error on Prediki’s platform wahlfieber.com for example, only amounted to 1.87 percentage points, the average error of all German political stock markets came to an excellent 2.65 percentage points.

Reliably more reliable

The applications show that this was not just an anecdotic coincidence. The political stock markets provided by far the best predictions for all three recent german elections. The difference was particularly striking in Saxony-Anhalt: the virtual exchange traders on wahlfieber.com only missed the official final result by 2.0 percentage points. This is 3 times more accurate when compared to traditional questionnaire surveys with the above mentioned error of 5.7 percentage points. The precision of political stock markets - which should not be confused with betting exchanges where gamblers risk real money - rests on several fundamentally new approaches. Respondents are not just used as voting cattle, rather they contribute their intelligence and analytical skills. In stark contrast to telephone surveys, political prediction markets do not suffer from selective response denial, nor false reports due to the social (un-)desirability of a political party.

Behavioural science has established, that paradoxically, people are much better in assessing the behaviour of others with less bias than their own behaviour. Political stock markets therefore do not ask: “How would you vote if the elections were on next Sunday?” they ask “How will people vote?”. Last but not least, traditional research agencies and the employees are invariably influenced by the media and the predictions of competing agencies. We now know that this was a main reason for the polling disaster before the UK parliamentary elections in 2015.

New capabilities for campaign managers

The last two years saw rapid progress in the capabilities of the technology. The most advanced political stock markets now can find out qualitative reasons to explain the quantitative predictions, using text analysis for the market talk of the virtual market participants. Traders continually try to convince others of their own investment opinions, to maximise their trading profits. Modern prediction markets decode these conversations for underlying sentiments and implicit arguments. The results are shown in real time dynamic charts.

Party strategists and campaign managers can now use this new tool to target their campaigns and events with more precision and higher efficiency. So far the only tool for this were focus groups which need comparatively much more lead time and are frequently unreliable due to dominant participants or group think. A better understanding of voters’ general will is not only a key puzzle piece to win elections, more importantly it is the best foundation for more political satisfaction of citizens and for better government.

• Back to the Management by Predictions Blog
• Follow this blog for more stories, new case studies and how-to's.